README.md

Small Scale Spatio-Temporal Forecasting Scenario

In this sub-scenario, we use and adapt the code repository of the LargeST repository.

1. Data Preparation

Follow the LargeST data format and handle all the benchmark datasets you want to use.

2. Experiments Running

To test Model (e.g. GWNet) w/o ST-TTC (i.e., Normal test) on different datasets (e.g. PEMS-08), you may execute the Python file in the terminal:

python experiments/gwnet/main.py --device cuda:1 --dataset PEMS08 --fewshot 100 --years 2016 --model_name gwnet --seed 2017 --bs 1 --mode test --eval_method norm

---

To test Model (e.g. GWNet) w/ ST-TTC on different datasets (e.g. PEMS-08), you may execute the Python file in the terminal:

python experiments/gwnet/main.py --device cuda:1 --dataset PEMS08 --fewshot 100 --years 2016 --model_name gwnet --seed 2017 --bs 1 --mode test --eval_method ttc